link to my thesis
link to my thesis
link to my thesis
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
232 CHAPTER 12. CONCLUSIONS & DIRECTIONS FOR FURTHER RESEARCH<br />
Furthermore, for lower complexities the required overhead for multiple processors<br />
dominates the actual computational costs and therefore a parallel approach requires<br />
more processing time than a single processor implementation.<br />
Finally, Section 7.5 describes four experiments in which the newly developed and<br />
implemented JDCOMM method is used <strong>to</strong> compute the global minima. The global<br />
minima in this section are also computed by the JD method and by the SOSTOOLS<br />
and GloptiPoly packages <strong>to</strong> be able <strong>to</strong> compare the performance and the efficiency<br />
of the various methods. The result here is that the JDCOMM method has a superior<br />
performance over the JD method and the SOSTOOLS and GloptiPoly methods for all<br />
the four experiments. The JDCOMM method uses fewer matrix-vec<strong>to</strong>r operations and<br />
therefore also needs less computation time.<br />
The performance of the iterative eigenvalue solvers is only efficient when the parameter<br />
settings are well chosen. They are important <strong>to</strong> ensure the convergence <strong>to</strong><br />
the desired eigenvalue. We have obtained additional insight on these settings. In<br />
particular, it is required <strong>to</strong> threshold the imaginary part of the target <strong>to</strong> prevent<br />
constant switching of target due <strong>to</strong> the approximations.<br />
In Part III of this <strong>thesis</strong> the approach of [50] plays an important role. In that<br />
article the H 2 model-order reduction problem from order N <strong>to</strong>t N − 1 is solved by<br />
algebraic methods. By approaching the problem as a rational optimization problem it<br />
is reformulated as the problem of finding solutions of a quadratic system of polynomial<br />
equations. In Chapter 8 this approach is generalized and extended which results in a<br />
joint framework in which <strong>to</strong> study the H 2 model-order reduction problem for various<br />
co-orders k ≥ 1. This makes it possible <strong>to</strong> solve the H 2 model-order reduction problem<br />
from order N <strong>to</strong> N − 1, N − 2, and N − 3.<br />
For co-orders k ≥ 1 we also end up with a system of equations in k − 1 additional<br />
real parameters. Furthermore, any feasible solution of this system of equations should<br />
satisfy k − 1 additional linear constraints. For computing the real approximations<br />
of order N − k, the algebraic Stetter-Möller matrix method for solving systems of<br />
polynomial equations of Section 3.3 is used. This approach casts the problem in<strong>to</strong> a<br />
particular eigenvalue problem involving one or more matrices in the k −1 parameters.<br />
This approach guarantees also in this application that the globally best approximation<br />
<strong>to</strong> the given system is found (while typically many local optima exist). To select the<br />
global best solution the new and useful result of Theorem 8.2 is used where we present<br />
for the co-order k case the homogeneous polynomial of degree three which coincides<br />
with the H 2 model-order reduction criterion for the difference between H(s) and G(s)<br />
at the stationary points of the system of equations. This is a generalization of the<br />
third order polynomial for the co-order k = 1 case presented in [50].<br />
In Chapter 9 this approach for k = 1 leads <strong>to</strong> a large conventional eigenvalue<br />
problem. The solutions of this eigenvalue problem lead <strong>to</strong> approximations of order<br />
N − 1 as described in [50]. However, we improve the efficiency of this method by a<br />
matrix-free implementation which is possible by using the nD-systems approach as<br />
presented in Chapter 5 in combination with an iterative Jacobi–Davidson eigenvalue